The field of program evaluation has expanded rapidly over the past three decades so that evaluation of programs is standard practice in many government agencies. One development is that evaluation activities have moved inside large government agencies with the establishment of internal evaluation staffs, procedures and policies. This study examined empirically the development of evaluation issues and processes in one branch of the National Science Foundation in the USA. The study, which involved 3 years of intermittent participant-observation and 44 interviews, indicates that evaluations (in this office) are heavily influenced by the way the work is organized and produced, as well as by the usual considerations that shape evaluations elsewhere. In particular, this paper addresses two primary issues identified in the study as critical to the successful establishment of an evaluation office: (I) the evolution of an evaluation culture within the organization, and (2) the management of a semi-internal, semi-external evaluation production process.